From: "Ville Syrjälä" <ville.syrjala@linux.intel.com>
To: "Lisovskiy, Stanislav" <stanislav.lisovskiy@intel.com>
Cc: intel-gfx@lists.freedesktop.org
Subject: Re: [Intel-gfx] [PATCH 5/9] drm/i915: Split skl+ plane update into noarm+arm pair
Date: Thu, 28 Oct 2021 16:59:28 +0300 [thread overview]
Message-ID: <YXqswK8DWDcfTA9B@intel.com> (raw)
In-Reply-To: <20211028135419.GA23664@intel.com>
On Thu, Oct 28, 2021 at 04:54:19PM +0300, Lisovskiy, Stanislav wrote:
> On Thu, Oct 28, 2021 at 04:03:32PM +0300, Ville Syrjälä wrote:
> > On Wed, Oct 27, 2021 at 08:11:37PM +0300, Lisovskiy, Stanislav wrote:
> > > On Mon, Oct 18, 2021 at 02:50:26PM +0300, Ville Syrjala wrote:
> > > > From: Ville Syrjälä <ville.syrjala@linux.intel.com>
> > > >
> > > > Chop skl_program_plane() into two halves. Fist half becomes
> > > > the _noarm() variant, second part the _arm() variant.
> > > >
> > > > Fortunately I have already previously grouped the register
> > > > writes into roughtly the correct order, so the split looks
> > > > surprisingly clean.
> > > >
> > > > A few notable oddities I did not realize were self arming
> > > > are AUX_DIST and COLOR_CTL.
> > > >
> > > > i915_update_info doesn't look too terrible on my cfl running
> > > > kms_atomic_transition --r plane-all-transition --extended:
> > > > w/o patch w/ patch
> > > > Updates: 2178 Updates: 2018
> > > > | |
> > > > 1us | 1us |
> > > > | |
> > > > 4us | 4us |*****
> > > > |********* |**********
> > > > 16us |********** 16us |*******
> > > > |*** |
> > > > 66us | 66us |
> > > > | |
> > > > 262us | 262us |
> > > > | |
> > > > 1ms | 1ms |
> > > > | |
> > > > 4ms | 4ms |
> > > > | |
> > > > 17ms | 17ms |
> > > > | |
> > > > Min update: 8332ns Min update: 6164ns
> > > > Max update: 48758ns Max update: 31808ns
> > > > Average update: 19959ns Average update: 13159ns
> > > > Overruns > 100us: 0 Overruns > 100us: 0
> > > >
> > > > And with lockdep enabled:
> > > > w/o patch w/ patch
> > > > Updates: 2177 Updates: 2172
> > > > | |
> > > > 1us | 1us |
> > > > | |
> > > > 4us | 4us |
> > > > |******* |*********
> > > > 16us |********** 16us |**********
> > > > |******* |*
> > > > 66us | 66us |
> > > > | |
> > > > 262us | 262us |
> > > > | |
> > > > 1ms | 1ms |
> > > > | |
> > > > 4ms | 4ms |
> > > > | |
> > > > 17ms | 17ms |
> > > > | |
> > > > Min update: 12645ns Min update: 9980ns
> > > > Max update: 50153ns Max update: 33533ns
> > > > Average update: 25337ns Average update: 18245ns
> > > > Overruns > 250us: 0 Overruns > 250us: 0
> > > >
> > > > TODO: On icl+ everything seems to be armed by PLANE_SURF, so we
> > > > can optimize this even further on modern platforms. But I
> > > > think there's a bit of refactoring to be done first to
> > > > figure out the best way to go about it (eg. just reusing
> > > > the current skl+ functions, or doing a lower level split).
> > > >
> > > > TODO: Split scaler programming as well, but IIRC the scaler
> > > > has some oddball double buffering behaviour on some
> > > > platforms, so needs proper reverse engineering
> > > >
> > > > Cc: Stanislav Lisovskiy <stanislav.lisovskiy@intel.com>
> > > > Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
> > > > ---
> > > > .../drm/i915/display/skl_universal_plane.c | 113 +++++++++++-------
> > > > 1 file changed, 72 insertions(+), 41 deletions(-)
> > > >
> > > > diff --git a/drivers/gpu/drm/i915/display/skl_universal_plane.c b/drivers/gpu/drm/i915/display/skl_universal_plane.c
> > > > index 74f3870d39b1..2a822e1e465e 100644
> > > > --- a/drivers/gpu/drm/i915/display/skl_universal_plane.c
> > > > +++ b/drivers/gpu/drm/i915/display/skl_universal_plane.c
> > > > @@ -1050,60 +1050,32 @@ static void icl_plane_csc_load_black(struct intel_plane *plane)
> > > > }
> > > >
> > > > static void
> > > > -skl_program_plane(struct intel_plane *plane,
> > > > - const struct intel_crtc_state *crtc_state,
> > > > - const struct intel_plane_state *plane_state,
> > > > - int color_plane)
> > > > +skl_program_plane_noarm(struct intel_plane *plane,
> > > > + const struct intel_crtc_state *crtc_state,
> > > > + const struct intel_plane_state *plane_state,
> > > > + int color_plane)
> > > > {
> > > > struct drm_i915_private *dev_priv = to_i915(plane->base.dev);
> > > > enum plane_id plane_id = plane->id;
> > > > enum pipe pipe = plane->pipe;
> > > > - const struct drm_intel_sprite_colorkey *key = &plane_state->ckey;
> > > > u32 stride = skl_plane_stride(plane_state, color_plane);
> > > > const struct drm_framebuffer *fb = plane_state->hw.fb;
> > > > - int aux_plane = skl_main_to_aux_plane(fb, color_plane);
> > > > int crtc_x = plane_state->uapi.dst.x1;
> > > > int crtc_y = plane_state->uapi.dst.y1;
> > > > - u32 x = plane_state->view.color_plane[color_plane].x;
> > > > - u32 y = plane_state->view.color_plane[color_plane].y;
> > > > u32 src_w = drm_rect_width(&plane_state->uapi.src) >> 16;
> > > > u32 src_h = drm_rect_height(&plane_state->uapi.src) >> 16;
> > > > - u8 alpha = plane_state->hw.alpha >> 8;
> > > > - u32 plane_color_ctl = 0, aux_dist = 0;
> > > > unsigned long irqflags;
> > > > - u32 keymsk, keymax;
> > > > - u32 plane_ctl = plane_state->ctl;
> > > > -
> > > > - plane_ctl |= skl_plane_ctl_crtc(crtc_state);
> > > > -
> > > > - if (DISPLAY_VER(dev_priv) >= 10)
> > > > - plane_color_ctl = plane_state->color_ctl |
> > > > - glk_plane_color_ctl_crtc(crtc_state);
> > > >
> > > > /* Sizes are 0 based */
> > > > src_w--;
> > > > src_h--;
> > > >
> > > > - keymax = (key->max_value & 0xffffff) | PLANE_KEYMAX_ALPHA(alpha);
> > > > -
> > > > - keymsk = key->channel_mask & 0x7ffffff;
> > > > - if (alpha < 0xff)
> > > > - keymsk |= PLANE_KEYMSK_ALPHA_ENABLE;
> > > > -
> > > > /* The scaler will handle the output position */
> > > > if (plane_state->scaler_id >= 0) {
> > > > crtc_x = 0;
> > > > crtc_y = 0;
> > > > }
> > > >
> > > > - if (aux_plane) {
> > > > - aux_dist = skl_surf_address(plane_state, aux_plane) -
> > > > - skl_surf_address(plane_state, color_plane);
> > > > -
> > > > - if (DISPLAY_VER(dev_priv) < 12)
> > > > - aux_dist |= skl_plane_stride(plane_state, aux_plane);
> > > > - }
> > > > -
> > > > spin_lock_irqsave(&dev_priv->uncore.lock, irqflags);
> > > >
> > > > /*
> > > > @@ -1119,16 +1091,10 @@ skl_program_plane(struct intel_plane *plane,
> > > > intel_de_write_fw(dev_priv, PLANE_SIZE(pipe, plane_id),
> > > > (src_h << 16) | src_w);
> > > >
> > > > - intel_de_write_fw(dev_priv, PLANE_AUX_DIST(pipe, plane_id), aux_dist);
> > > > -
> > > > if (icl_is_hdr_plane(dev_priv, plane_id))
> > > > intel_de_write_fw(dev_priv, PLANE_CUS_CTL(pipe, plane_id),
> > > > plane_state->cus_ctl);
> > > >
> > > > - if (DISPLAY_VER(dev_priv) >= 10)
> > > > - intel_de_write_fw(dev_priv, PLANE_COLOR_CTL(pipe, plane_id),
> > > > - plane_color_ctl);
> > > > -
> > > > if (fb->format->is_yuv && icl_is_hdr_plane(dev_priv, plane_id))
> > > > icl_program_input_csc(plane, crtc_state, plane_state);
> > > >
> > > > @@ -1138,6 +1104,52 @@ skl_program_plane(struct intel_plane *plane,
> > > >
> > > > skl_write_plane_wm(plane, crtc_state);
> > > >
> > > > + intel_psr2_program_plane_sel_fetch(plane, crtc_state, plane_state, color_plane);
> > > > +
> > > > + spin_unlock_irqrestore(&dev_priv->uncore.lock, irqflags);
> > > > +}
> > > > +
> > > > +static void
> > > > +skl_program_plane_arm(struct intel_plane *plane,
> > > > + const struct intel_crtc_state *crtc_state,
> > > > + const struct intel_plane_state *plane_state,
> > > > + int color_plane)
> > > > +{
> > > > + struct drm_i915_private *dev_priv = to_i915(plane->base.dev);
> > > > + enum plane_id plane_id = plane->id;
> > > > + enum pipe pipe = plane->pipe;
> > > > + const struct drm_intel_sprite_colorkey *key = &plane_state->ckey;
> > > > + const struct drm_framebuffer *fb = plane_state->hw.fb;
> > > > + int aux_plane = skl_main_to_aux_plane(fb, color_plane);
> > > > + u32 x = plane_state->view.color_plane[color_plane].x;
> > > > + u32 y = plane_state->view.color_plane[color_plane].y;
> > > > + u32 keymsk, keymax, aux_dist = 0, plane_color_ctl = 0;
> > > > + u8 alpha = plane_state->hw.alpha >> 8;
> > > > + u32 plane_ctl = plane_state->ctl;
> > > > + unsigned long irqflags;
> > > > +
> > > > + plane_ctl |= skl_plane_ctl_crtc(crtc_state);
> > > > +
> > > > + if (DISPLAY_VER(dev_priv) >= 10)
> > > > + plane_color_ctl = plane_state->color_ctl |
> > > > + glk_plane_color_ctl_crtc(crtc_state);
> > > > +
> > > > + keymax = (key->max_value & 0xffffff) | PLANE_KEYMAX_ALPHA(alpha);
> > > > +
> > > > + keymsk = key->channel_mask & 0x7ffffff;
> > > > + if (alpha < 0xff)
> > > > + keymsk |= PLANE_KEYMSK_ALPHA_ENABLE;
> > > > +
> > > > + if (aux_plane) {
> > > > + aux_dist = skl_surf_address(plane_state, aux_plane) -
> > > > + skl_surf_address(plane_state, color_plane);
> > > > +
> > > > + if (DISPLAY_VER(dev_priv) < 12)
> > > > + aux_dist |= skl_plane_stride(plane_state, aux_plane);
> > > > + }
> > > > +
> > > > + spin_lock_irqsave(&dev_priv->uncore.lock, irqflags);
> > > > +
> > > > intel_de_write_fw(dev_priv, PLANE_KEYVAL(pipe, plane_id),
> > > > key->min_value);
> > > > intel_de_write_fw(dev_priv, PLANE_KEYMSK(pipe, plane_id), keymsk);
> > > > @@ -1146,17 +1158,22 @@ skl_program_plane(struct intel_plane *plane,
> > > > intel_de_write_fw(dev_priv, PLANE_OFFSET(pipe, plane_id),
> > > > (y << 16) | x);
> > >
> > > Wondering, if we could also move PLANE_OFFSET register to non-arming
> > > section.
> >
> > No. It is self arming.
> >
> > > BSpec states, that it is not arming and being armed by
> > > writing PLANE_SURF. I think we even agreed to check if that was true,
> > > so I made some tests with DG2, I think and writes to PLANE_OFFSET
> > > didn't anyhow affect the hw until PLANE_SURF was written(was checking
> > > it by reading the crcs...)
> >
> > Your test was invalid since it was done on icl+ where everything is
> > armed by PLANE_SURF.
>
> Well, at that moment the main question was about DG2/ADL, so I tested
> it with that particular platform(we had atomic update issues mostly there)
>
> Also I guess it means PLANE_OFFSET can be moved to non-arming section
> for icl+ at least.
>
> Or was it done somewhere later in your series?
> Didn't walk through all of the patches yet.
Quoting from the commit message of this very patch:
"TODO: On icl+ everything seems to be armed by PLANE_SURF, so we
can optimize this even further on modern platforms. But I
think there's a bit of refactoring to be done first to
figure out the best way to go about it (eg. just reusing
the current skl+ functions, or doing a lower level split). "
--
Ville Syrjälä
Intel
next prev parent reply other threads:[~2021-10-28 13:59 UTC|newest]
Thread overview: 38+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-10-18 11:50 [Intel-gfx] [PATCH 0/9] drm/i915: Split plane updates to noarm+arm phases Ville Syrjala
2021-10-18 11:50 ` [Intel-gfx] [PATCH 1/9] drm/i915: Reject planar formats when doing async flips Ville Syrjala
2021-10-27 16:12 ` Lisovskiy, Stanislav
2021-10-18 11:50 ` [Intel-gfx] [PATCH 2/9] drm/i915: Fix async flip with decryption and/or DPT Ville Syrjala
2021-11-03 18:39 ` Lisovskiy, Stanislav
2021-10-18 11:50 ` [Intel-gfx] [PATCH 3/9] drm/i915: Fix up the sprite namespacing Ville Syrjala
2021-11-03 18:47 ` Lisovskiy, Stanislav
2021-10-18 11:50 ` [Intel-gfx] [PATCH 4/9] drm/i915: Split update_plane() into update_noarm() + update_arm() Ville Syrjala
2021-10-27 16:35 ` Lisovskiy, Stanislav
2021-11-03 18:47 ` Lisovskiy, Stanislav
2021-10-18 11:50 ` [Intel-gfx] [PATCH 5/9] drm/i915: Split skl+ plane update into noarm+arm pair Ville Syrjala
2021-10-18 12:06 ` Lisovskiy, Stanislav
2021-10-18 17:14 ` Ville Syrjälä
2021-10-18 17:22 ` Ville Syrjälä
2021-10-27 17:11 ` Lisovskiy, Stanislav
2021-10-28 13:03 ` Ville Syrjälä
2021-10-28 13:54 ` Lisovskiy, Stanislav
2021-10-28 13:59 ` Ville Syrjälä [this message]
2021-10-28 14:05 ` Lisovskiy, Stanislav
2021-11-03 18:46 ` Lisovskiy, Stanislav
2021-10-18 11:50 ` [Intel-gfx] [PATCH 6/9] drm/i915: Split pre-skl primary " Ville Syrjala
2021-10-20 21:27 ` [Intel-gfx] [PATCH v2 " Ville Syrjala
2021-11-03 18:49 ` Lisovskiy, Stanislav
2021-11-03 18:47 ` [Intel-gfx] [PATCH " Lisovskiy, Stanislav
2021-10-18 11:50 ` [Intel-gfx] [PATCH 7/9] drm/i915: Split g4x+ sprite " Ville Syrjala
2021-11-03 18:46 ` Lisovskiy, Stanislav
2021-10-18 11:50 ` [Intel-gfx] [PATCH 8/9] drm/i915: Split ivb+ " Ville Syrjala
2021-11-03 18:45 ` Lisovskiy, Stanislav
2021-10-18 11:50 ` [Intel-gfx] [PATCH 9/9] drm/i915: Split vlv/chv " Ville Syrjala
2021-11-03 18:44 ` Lisovskiy, Stanislav
2021-10-18 13:06 ` [Intel-gfx] ✗ Fi.CI.CHECKPATCH: warning for drm/i915: Split plane updates to noarm+arm phases Patchwork
2021-10-18 13:08 ` [Intel-gfx] ✗ Fi.CI.SPARSE: " Patchwork
2021-10-18 13:38 ` [Intel-gfx] ✓ Fi.CI.BAT: success " Patchwork
2021-10-18 16:43 ` [Intel-gfx] ✗ Fi.CI.IGT: failure " Patchwork
2021-10-20 22:39 ` [Intel-gfx] ✗ Fi.CI.CHECKPATCH: warning for drm/i915: Split plane updates to noarm+arm phases (rev2) Patchwork
2021-10-20 22:40 ` [Intel-gfx] ✗ Fi.CI.SPARSE: " Patchwork
2021-10-20 23:08 ` [Intel-gfx] ✓ Fi.CI.BAT: success " Patchwork
2021-10-21 3:19 ` [Intel-gfx] ✓ Fi.CI.IGT: " Patchwork
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=YXqswK8DWDcfTA9B@intel.com \
--to=ville.syrjala@linux.intel.com \
--cc=intel-gfx@lists.freedesktop.org \
--cc=stanislav.lisovskiy@intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox